meta-matrix-skill-invocation

You can validate Claude Code skill invocation paths by running end-to-end probes that test direct slash-commands, the Skill tool, and subagent scenarios to confirm consistent argument propagation and execution.

$ARGUMENTS propagation in Claude Code ensures that variables pass correctly across direct commands, mid-run Skill tool calls, and fork or preload subagents without losing context or gating execution.

Mid-run Skill tool failures in subagent contexts often stem from unvalidated invocation paths where fork or preload boundaries block $ARGUMENTS propagation, requiring end-to-end path testing to identify gating prompts.

Claude Code supports both direct slash-command and preload subagent invocation, but you must validate all four skill invocation paths end-to-end to ensure consistent results and argument propagation across agent boundaries.

Test skill orchestration reliability by deploying internal probes to verify sentinel signaling and path execution across sub-skills, confirming that direct commands, Skill tool calls, and fork subagents behave consistently.

Fork subagent skill invocation can encounter gating prompts and inconsistent path execution if not validated end-to-end, making it necessary to probe $ARGUMENTS propagation and sentinel signaling to ensure reliable behavior.